基于小波神经网络的地铁轴承故障诊断方法

New Wavelet Neural Network Method for Fault Diagnosis of Metro Bearings

提出了一种将小波包变换和改进BP神经网络相结合的地铁轴承故障诊断模型。该模型采用小波包变换对采集到的原始振动信号进行分解和重构,提取能量特征向量,并采用一种改进的小波神经网络模型对地铁轴承进行故障诊断,引入动量因子优化BP神经网络梯度下降算法,以提高模型诊断精度和收敛速度。基于凯斯西储大学轴承故障实验中加速度传感器采集的数据,首先对其进行预处理,然后对神经网络进行训练,当训练误差达到目标精度时,利用该模型开展地铁轴承故障诊断仿真实验。结果表明:改进的小波神经网络模型可以快速、准确地诊断出地铁轴承的故障类型。

A fault diagnosis model of metro bearings was proposed,which combines wavelet packet transform with improved BP neural network.To reduce the environmental noise,this model uses wavelet packet transform to decompose and reconstruct the collected original vibration signals,and extract the energy feature vectors.To ensure the diagnostic accuracy and convergence rate,the momentum factor was introduced to optimize BP neural network gradient descent algorithm.Based on the data collected by the acceleration sensor in the bearing fault experiment of Case Western Reserve University,the signals were preprocessed.Then,the neural network was trained.When the training error meets the requirement,the model was used to carry out the fault diagnosis simulation experiment of metro bearings.Experimental result shows that the improved wavelet neural network model can diagnose rapidly and precisely the fault types of metro bearings.